AI Paper Atlas

Research map / Architectures and efficiency / Spiking networks

Pseudo-Spiking Neurons: A Noise-Based Training Framework for Heterogeneous-Latency Spiking Neural Networks

Yuxuan Zhang, Yuhang Sun, Hongjue Li, Yue Deng, Wen Yao

AAAI 2026 · Architectures and efficiency › Spiking networks · approach: Training deep spiking networks · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
learn heterogeneous per-layer timesteps with noise proxy instead of fixed timestepSNN ResNet-19 · CIFAR-100top-1 accuracy: 80.71 -> 81.29 vs RateBP at ~3 avg timesteps (MTT 81.51 at T=4)Mixed
learned heterogeneous latency vs RateBP on ImageNetSEW-ResNet-34 · ImageNettop-1 accuracy: 65.84 -> 68.18Improved
add both quantization and clipping noise (vs single term) (ablation)PseudoSNN ResNet-18 · CIFAR-100top-1 accuracy: 31.16 -> 79.02 (clipping only vs full, no calibration)Improved
add 10-epoch calibration fine-tuning (ablation)PseudoSNN ResNet-18 · CIFAR-100top-1 accuracy: 79.02 -> 79.57Improved
use PseudoSN proxy vs BPTT/RateBP for trainingResNet-18 · CIFAR-100 training costtraining time and memoryImproved
learned heterogeneous timesteps vs uniform timestep SEW-ResNet-34SEW-ResNet-34 · ImageNetenergy (mJ): 4.89 -> 4.13 mJImproved
vary lambda in hardware-aware penalty (ablation)PseudoSNN ResNet-18 · CIFAR-100accuracy-latency ParetoImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy